In the realm of Silicon Valley’s tech discussions, the spotlight has been firmly fixed on Generative Artificial Intelligence (AI). This technology has dazzled us with its human-like ability to draw inferences through the use of Large Language Models (LLMs). However, there’s another exciting development gaining momentum – the convergence of AI with the Internet of Things (IoT).
**Edge** computing, often used interchangeably with **IoT**, demands clarification. The IoT represents the ecosystem of interconnected devices, while edge computing describes the processing that occurs on these devices. IoT devices typically require internet connectivity to function, whereas edge devices may operate in a disconnected state for extended periods, only occasionally connecting to a cloud data center for processing.
Creating hardware for edge applications necessitates a unique design, considering factors like computational performance, power efficiency, and economic viability. The IT industry has been diligently working to enhance edge computing capabilities, making it sharper and more efficient.
The expanding IoT universe
By 2030, it’s anticipated that over 125 billion IoT devices will be connected to the Internet, ranging from smartphones and cameras to smart home appliances. These devices will generate massive amounts of data, with a staggering 80% of it comprising video and images. Despite the abundance of data, only a small fraction has been analyzed so far.
Growing concerns surrounding privacy, security, and bandwidth limitations have prompted a shift towards processing data closer to its source, at the edge of the IoT. Can AI come to the rescue in this context? Currently, AI technologies are primarily designed for cloud computing operations, which don’t face the same constraints as edge devices. This is where AI specialist Axelera AI steps in.
Axelera AI’s Metis AI platform
Axelera AI has unveiled its Metis AI Platform, which has recently entered the early access phase for developing advanced edge AI-native hardware and software solutions. According to Fabrizio Del Maffeo, Axelera AI’s co-founder and CEO, “Placing a comprehensive hardware and software solution directly into the hands of our customers within a mere 25 months stands as a pivotal milestone for our company.”
The Metis AI Platform offers practical edge AI inference solutions, tailored to companies developing next-generation computer vision applications. It simplifies real-world deployment, providing a user-friendly path to development and integration. Available in industry-standard form factors such as PCIe cards and vision-ready systems, it streamlines the integration of AI into business applications.
Unveiling dataflow technology
At the heart of the Metis AI Platform lies the Metis AI Processing Unit (AIPU), based on proprietary digital in-memory computing technology (D-IMC) and RISC-V with ‘dataflow’ technology. Dataflow computers prioritize optimizing data movement within applications and harness massive parallelism between numerous ‘dataflow cores,’ delivering significant performance, space, and power efficiency improvements.
Axelera AI claims that the AIPU offers industry-leading performance, usability, and efficiency at a fraction of the cost of existing solutions. Furthermore, the technology is scalable for deployment in projects experiencing growth, and it incorporates an embedded security engine to protect sensitive data through encryption, ensuring the security of biometric information.
Empowering small and medium-sized enterprises
Axelera AI’s technology is integrated into AI acceleration cards, AI acceleration boards, and AI acceleration vision-ready systems, making it accessible to small and medium-sized enterprises. This approach expedites adoption and simplifies field installation. To further facilitate development, Axelera AI offers the Voyager Software Development Kit (SDK).
The Voyager SDK offers a streamlined way for developers to build high-performance applications for the Metis AI platform. Developers can describe their end-to-end pipelines declaratively using a simple YAML configuration file, including deep learning models and non-neural pre and post-processing elements. The SDK toolchain automates model compilation and deployment for the Metis AI platform, efficiently allocating processing components to available computing elements.
The future of edge computing
As the world becomes increasingly reliant on IoT devices, it’s crucial to harness the potential of edge computing. Axelera AI’s Metis AI Platform represents a significant step forward, offering efficient AI solutions that cater to the unique challenges of edge devices. With AI becoming more accessible and adaptable, we are on the cusp of a transformative era where even the simplest devices, like smart toasters, can benefit from advanced machine learning capabilities.
The edge is becoming smarter, sharper, and more expansive. It’s imperative that we maintain control over this emerging realm of device intelligence to ensure it remains a force for good, rather than taking a darker turn. With innovations like Axelera AI’s Metis AI Platform, we are poised to unlock the tangible results that will define the next phase of AI-driven technology.
